The OSW products are two-dimensional ocean surface wave spectra estimated from a Level-1 SLC by inversion of the corresponding image cross-spectra. The cross-spectra are computed by performing multi-looking in azimuth followed by co- and cross-spectra estimation among the detected individual look images. A single OSW can be computed from an Wave mode SLC imagette or from a sub-image extracted from a Stripmap mode SLC image.

The OSW product cannot be generated from the TOPSAR mode, since individual looks with sufficient time separation are required. The obtained inter-look time separation within one burst is too short due to the progressive scanning (i.e. short dwell time). Individual looks from neighbouring bursts require significant spatial overlap.

The Sentinel-1 wave processing system consists of:

• a spectral estimation unit
• a spectral inversion unit

The OSW processing will use the following inputs:

• SM or WV SLC product
• External auxiliary data:
  • ECMWF atmospheric model data
  • Level-2 processor parameters auxiliary data
  • Simulated cross-spectra data
• Internal auxiliary data:
  • coastline and land masking data
  • General Bathymetry Chart of the Oceans (GEBCO)(in case of SM data)
  • Range Fourier profile
  • IPF Level-2 internal parameter file containing extra processing parameters specific to the OSW algorithm.

Spectral Estimation

The spectral estimation unit performs the processing from Level-1 SLC product to an internal co- and cross-variance function (Level-1B) product.

The spectral estimation consists of inter-look cross spectral processing based on splitting the azimuth bandwidth into three non-overlapping looks, followed by an estimation of the co- and cross-variance function based on the periodogram method.

The final result consists of one co-variance function and two cross-variance functions on cartesian grid. The two cross-variance functions correspond to the neigbour looks and the outer looks i.e. with two different look separation times. The co-variance function is the average of the co-variance functions from the three individual looks. The processing also estimates the percentage of land within the selected estimation area (in case of SM data), the range and azimuth cut-off wavelength, spectral resolution, and some image statistics (mean, variance, skewness).

The figure below shows the flowchart for the spectral estimation.

Spectral Inversion

The spectral inversion unit generates the Level-2 OSW product using the intermediate Level-1 data product as input.

The OSW spectral inversion unit first accesses the intermediate product and performs a 2D Fourier transform to achive the co- and cross-spectra on cartesian grid. A Hanning window is used in the 2D Fourier transform. The OSW processing then performs a wave spectral inversion of the co- and cross-spectra with respect to the detected SAR ocean wave-like pattern. This is done by first estimating and removing the non-linear contribution to the imaging process, assuming that this is caused only by the local wind field, and then applying a quasi-linear inversion in the most energetic part of the SAR co- and cross-spectrum. The wind field is therefore required, and this is estimated as described in the OSW processing.

The estimation of wind sea significant wave height is performed using the estimated wind speed and the azimuth cut-off wavelength.

The major requirements for the quality of the inversion is knowledge of the Real Aperture Radar (RAR) Modulation Transfer Function (MTF), the azimuth cut-off wavelength, and an accurate removal of the non-linear part of the spectra (i.e. the wind field). The RAR MTF is computed using a backscattering model including non-uniform distribution of scatterers on the long wave field. After the inversion, the ocean wave spectrum is converted to polar grid, rotated relative to north, partitioned, and ambiguity resolved followed by computation of key spectral parameters for the two most energetic partitions. Finally, an output product is generated from the polar spectra and stored in a netCDF format together with extracted parameters stored as attributes and some key parameters from the corresponding Level-1 product.

The figure below shows the flowchart for the spectral inversion.